---
license: apache-2.0
library_name: diffusers
pipeline_tag: image-to-image
tags:
  - controlnet
  - diffusion
  - stable-diffusion
  - ui-generation
  - conditional-generation
base_model: runwayml/stable-diffusion-v1-5
---

# ControlNet for Mobile UI Layout Generation

This repository contains a ControlNet model fine-tuned to generate stylized mobile user interface (UI) screens from wireframe layouts and structured text prompts.

The model was trained on a fully synthetic dataset of mobile UI layouts, allowing precise control over spatial structure and design parameters.

## Model Overview

- Architecture: ControlNet + Stable Diffusion 1.5
- Conditioning:
  - Wireframe image (layout constraints)
  - Text prompt (design parameters)
- Resolution: 512 × 512
- Training data: Procedurally generated synthetic UI layout

## Usage

This model is designed to be used with the Stable Diffusion ControlNet pipeline.

```python
import torch
from diffusers import ControlNetModel, StableDiffusionControlNetPipeline
from PIL import Image

# Load ControlNet
controlnet = ControlNetModel.from_pretrained(
    "louis-gs/controlnet-mobile-ui-layout",
    torch_dtype=torch.float16
)

# Load Stable Diffusion + ControlNet pipeline
pipe = StableDiffusionControlNetPipeline.from_pretrained(
    "runwayml/stable-diffusion-v1-5",
    controlnet=controlnet,
    torch_dtype=torch.float16,
    safety_checker=None
).to("cuda")

# Load conditioning image (wireframe)
conditioning_image = Image.open("wireframe.png").convert("RGB")

# Structured prompt (same format as training)
prompt = (
    "a light mobile app product screen UI, "
    "low density, primary color palette p5, "
    "rounded corners radius 8, "
    "topbar with_search, bottom navigation 3, "
    "tabs 2, hero carousel, "
    "1 cards, 6 list items, "
    "cta none, 1 badges, 2 sections, "
    "header_block true, price_tag true"
)

# Generate image
image = pipe(
    prompt=prompt,
    image=conditioning_image,
    num_inference_steps=30,
    guidance_scale=7.5
).images[0]

image.save("result.png")